Consider the general optimization problem
Suppose that the object function and the constraint functions
and
are continuously differentiable at a point
If
is a local optimum and optimization problem satisfies some conditions, then there exist constant
and
, called KKT multipliers, such that
KKT condition
- Stationarity:
- Primal feasibility:
- Dual feasibility:
Complementary slackness:
Note:
- KKT condition is a necessary condition.
- KKT condition is necessary and sufficient condition under Slater condition.
-
Slater condition: For a convex problem(i.e.
are convex and
is affine), there exists a point
such that
and
KKT.png